This proposal will test the efficacy of a promising pharmacological approach for the treatment of comorbid cocaine dependence and bipolar disorder. We propose a randomized, double blind, placebo-controlled, 12-week trial, to test the efficacy of Divalproex sodium, "valproate" + Treatment-As-Usual (TAU) compared to placebo+ TAU (TAU = lithium carbonate + supportive psychosocial treatment) in decreasing cocaine use and stabilizing mood symptoms among patients with comorbid cocaine dependence and bipolar disorder. This proposal addresses a major area of interest identified in NIDA's Division of Treatment Research and Development (DTRD) Research Programs on "Medications for the Treatment of Comorbid Alcohol and/or Mental Disorders and Drug Abuse." Bipolar disorder has the highest rate of association with cocaine and other substance use disorders (SUDs) than any other major severe psychiatric syndromes. This comorbidity represents a major treatment challenge and is associated with severe disability, morbidity, and heightened risk for suicide. Remarkably little research has been aimed at identifying optimal pharmacological treatments, and treatment needs for these patients remain substantially unmet. Our recently concluded double blind, placebo-controlled study of valproate in bipolar alcoholics remains the only trial completed to date addressing any bipolar-SUD comorbidity. Our work had shown a significant advantage for valproate over placebo on decreasing heavy drinking. There are compelling theoretical, and accruing clinical evidence suggesting that valproate may be effective in reducing cocaine use among this severely impaired population. A secondary analysis of the above mentioned valproate trial showed a significant advantage of valproate over placebo on decreasing cocaine use, including cocaine positive urine screens. Additionally, results from our open label pilot study of valproate for bipolar disorder with comorbid cocaine dependence, also showed a significant decrease in cocaine use, along with unproved mood and functioning. These promising findings are in agreement with findings of a number of open label studies testing the utility of valproate in cocaine and other SUDs. There is a need for evidence-based interventions to guide and inform treatment choices for these high-risk patients. The aims of this study are: 1) Examine the efficacy of valproate+TAU compared to placebo+TAU in decreasing cocaine use in patients with DSM-IV cocaine dependence and comorbid bipolar I disorder; 2) Determine whether primary vs. secondary cocaine dependence, bipolar subtype (depressed vs. manic/mixed), and the presence of additional SUDs moderate the association between treatment and cocaine use outcome; 3) Assess the effects of medication compliance and mood symptoms as mediators of cocaine use outcome.